A reference human genome sequence and the subsequent decade-long annotation of ~20,000 (20K) protein- coding genes has enabled an explosion of disease-associated genomic variant discovery. We can now anticipate a nearly full description of all disease-related genomic variations in the human population. Genomic sequencing, however, if performed in isolation, will leave fundamental questions about genotype-phenotype relationships unresolved. For the vast majority of genomic variations identified, it remains unclear if and how they perturb the function of the corresponding genes or gene products. To ?connect the dots? of the genomic revolution, functions and context must be assigned for large numbers of genotypic changes. Currently, we know relatively few of the molecular, biochemical, and functional interactions that take place in human cells and that are necessary for biological functions. Past discoveries about interactions such as protein-protein interactions (PPIs) have been highly biased towards pairs of `popular' proteins, representing a tiny fraction of the full space of 200,000,000 pairings of 20Kx20K genes. Systematic, high-quality, genome-wide efforts to create community resources of molecular interactions constitute the best solution to this problem. Just as reference genome sequences provided a fundamental community resource that revolutionized human genetics, reference maps of genome-wide or proteome-wide interaction networks, or ?interactome networks?, will be critical to fully understand genotype-phenotype relationships. This application is the fifth competitive renewal of a grant funded by NHGRI since 1998 to address the challenge described above through experimental mapping of binary interactome networks at proteome-scale. After devoting two cycles to the development of interactome mapping strategies in model organisms, this is the third renewal specifically addressing human PPIs. We are now at the exciting stage of presenting a three-year roadmap to deliver ?A human binary interactome reference map by 2018? as a broadly useful resource for the scientific community, with pre-publication release of 12 complementary high-quality genome-wide 20Kx20K PPI datasets along the way. Ultimately, the resulting reference map, which we estimate may well be an order of magnitude larger than the collective efforts of the scientific community to detect PPIs using small-scale experiments, will be an invaluable tool to connect the dots of genomics and will serve as a scaffold to initiate unbiased and exhaustive functional characterizations of large numbers of genomic variations associated with human disease.